Research map / Large language models / Agents, code and math
Forecasting Future World Events With Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add retrieval (FiD Static 2.8B) instead of fine-tuned T5 2.8B | T5 · Autocast | combined Score: 32.5 -> 40.6 | Helped |
| Add retrieval (FiD Static) instead of zero-shot UnifiedQA | UnifiedQA · Autocast | Average score: 19.5 -> 37.2 | Helped |
| Scale FiD Static from 0.2B to 2.8B | FiD Static · Autocast | combined Score: 33.5 -> 40.6 | Helped |
| Scale T5 from 0.2B to 2.8B | T5 · Autocast | combined Score: 32.4 -> 32.5 | No clear effect |
| Scale UnifiedQA from 0.2B to 2.8B | UnifiedQA · Autocast | combined Score: 17.2 -> 22.8 | Helped |
| Use FiD Temporal (crowd auxiliary supervision) instead of FiD Static | FiD Static · Autocast | Average score: 37.2 -> 37.8 | No clear effect |
| Use FiD Temporal instead of FiD Static, largest models | FiD · Autocast | combined Score: 40.6 -> 40.1 | Hurt |
| Use FiD Temporal instead of FiD Static, MCQ at largest size | FiD · Autocast | MCQ accuracy (%): 35.8 -> 36.9 | Helped |
| Use FiD Temporal instead of FiD Static, numerical at largest size | FiD · Autocast | Numerical l1 distance (%): 19.9 -> 19.5 | Helped |
| Use crowd predictions as auxiliary signal to reduce calibration error | FiD Temporal · Autocast resolved test set | calibration error: 40 -> 17 | Helped |
| Scale DeBERTa-v3 from 22M to 304M | DeBERTa-v3 · IntervalQA | RMS calibration error: 19.1 -> 13.5 | Helped |
| Scale DeBERTa-v3 from 22M to 304M | DeBERTa-v3 · IntervalQA | confidence interval length: 2072.4 -> 305.4 | Helped |
Closest papers by meaning
- AutoCast++: Enhancing World Event Prediction with Zero-shot Ranking-based Context Retrieval (ICLR 2024 · similarity 0.91)
- Approaching Human-Level Forecasting with Language Models (NeurIPS 2024 · similarity 0.86)
- AutoTimes: Autoregressive Time Series Forecasters via Large Language Models (NeurIPS 2024 · similarity 0.84)
- Context is Key: A Benchmark for Forecasting with Essential Textual Information (ICML 2025 · similarity 0.84)
- GPT4MTS: Prompt-based Large Language Model for Multimodal Time-series Forecasting (AAAI 2024 · similarity 0.83)
- A decoder-only foundation model for time-series forecasting (ICML 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.